Professional Documents
Culture Documents
Understanding and Eliminating EMI in Microcontroller Applications
Understanding and Eliminating EMI in Microcontroller Applications
http://www.national.com 2
3.2 ElectroMagnetic Compatibility car causing interference in the radio. This type of interfer-
EMI or electrical noise is of world wide concern. Govern- ence is more difficult to contain because, as mentioned ear-
mental agencies and certain industry bodies have issued lier, the systems are generally not designed by a single per-
specifications with which all electrical, electromechanical, son. However, design methods and control techniques used
and electronic equipment must comply. These specifica- to contain the intra-system form of EMI, which are almost
tions and limitations are an attempt to ensure that proper always under the control of a single user, will inherently help
EMC techniques are followed by manufacturers during the reduce the inter-system noise.
design and fabrication of their products. When these tech- This paper will address problems and solutions in the area
niques are properly applied, the product can then operate of intra-system noise. Intra-system interference situations
and perform with other equipment in a common environ- occur when the sources, victims, and coupling paths are
ment so that no degradation of performance exists due to entirely within one system, module or PCB. Systems may
internally or externally conducted or radiated electromag- provide emissions that are conducted out of power lines or
netic emissions. This is defined as ElectroMagnetic Com- be susceptible to emissions conducted through them. Sys-
patibility or EMC. tems may radiate emissions through space in addition to
being susceptible to radiated noise. Noise conducted out
antenna leads turns into radiated noise. By the same token,
radiated noise picked up by the antenna is turned into con-
ducted noise within the system. A perfect example is a PCB
ground loop, which makes an excellent antenna. The sys-
tem itself is capable of degrading its own performance due
to its own internal generation of conducted and radiated
noise and its susceptibility to it.
Some results of EMI within a system: Noise on power lines
causing false triggering of logic circuits, rapidly changing
signals causing ‘‘glitches’’ on adjacent steady state signal
lines (crosstalk) causing erratic operation, multiple simulta-
neously switching logic outputs propagating ground bounce
noise throughout system, etc.
3.4 Coupling Paths
The modes of coupling an emitter source to a receptor vic-
tim can become very complicated. Remember, each EMI
situation can be classified into two categories of coupling,
TL/DD/12857 – 1
conducted and radiated. Coupling can also result from a
FIGURE 1. EMI Situation
combination of paths. Noise can be conducted from an
3.3 Inter-System EMI emitter to a point of radiation at the source antenna, then
For the purpose of this paper, when the source of noise is a picked up at the receptor antenna by induction, and re-con-
module, board, or system and the victim is a different and ducted to the victim. A further complication that multiple
separate module, board, or system under the design or con- coupling paths presents is that it makes it difficult to deter-
trol of a different person, it will be considered an inter-sys- mine if eliminating a suspected path has actually done any
tem interference situation. Examples of inter-system inter- good. If two or more paths contribute equally to the prob-
ference situations could be a Personal Computer interfering lem, eliminating only one path may provide little apparent
with the operation of a TV or an anti-lock brake module in a improvement.
TL/DD/12857 – 2
FIGURE 2. Intra-System EMI Manifestations
3 http://www.national.com
3.5 Conducted Interference
In order to discuss the various ways in which EMI can cou-
ple from one system to another, it is necessary to define a
few terms. There are two varieties of conducted interfer-
ence which concern us. The first variety is differential-mode
interference. That is an interference signal that appears be-
tween the input terminals of a circuit. The other variety of
conducted interference is called common-mode interfer-
ence. A common-mode interference signal appears be-
TL/DD/12857 – 4
tween each input terminal and a third point, the common-
FIGURE 4. Differential-Mode Interference
mode reference. That reference may be the equipment
chassis, an earth ground, or some other point.
Let’s look at each type of interference individually. In Figure
3 we show a simple circuit consisting of a signal source, VS,
and a load, RL. In Figure 4 we show what happens when
differential-mode interference is introduced into the circuit
by an outside source. As is shown, an interference voltage,
VD, appears between the two input terminals, and an inter-
ference current, ID, flows in the circuit. The result is noise at
the load. If, for instance, the load is a logic gate in a comput-
er, and the amplitude of VD is sufficiently high, it is possible
for the gate to incorrectly change states.
Figure 5 shows what happens when a ground loop is added
to our circuit. Ground loops, which are undesirable current
paths through a grounded body (such as a chassis), are TL/DD/12857 – 5
usually caused by poor design or by the failure of some FIGURE 5. Common-Mode Interference
component. In the presence of an interference source, com- 3.6 Radiated Interference
mon-mode currents, IC, and a common-mode voltage, VC,
Radiated coupling itself can take place in one of several
can develop, with the ground loop acting as the common-
ways. Some of those include field-to-cable coupling, cable-
mode reference. The common-mode current flows on both
to-cable coupling, and common-mode impedance coupling.
input lines, and has the same instantaneous polarity and
Let’s look at those types of coupling one at a time.
direction (the current and voltage are in phase), and returns
through the common-mode reference. The common-mode The principle behind field-to-cable coupling is the same as
voltage between each input and the common-mode refer- that behind the receiving antenna. That is, when a conduc-
ence is identical. tor is placed in a time-varying electromagnetic field, a cur-
rent is induced in that conductor. That is shown in Figure 6 .
In this figure, we see a signal source, VS, driving a load, RL.
Nearby there is a current carrying wire (or other conductor).
Surrounding the wire is an electromagnetic field induced by
the current flowing in the wire. The circuit acts like a loop
antenna in the presence of this field. As such, an interfer-
ence current, IN, and an interference voltage, VN, are in-
duced in the circuit. The magnitude of the induced interfer-
TL/DD/12857–3
ence signal is roughly proportional to the magnitude of the
FIGURE 3. Simple Circuit
incoming field, the frequency of the incoming field, the size
of the loop, and the impedance of the loop.
TL/DD/12857 – 6
FIGURE 6. Field-to-Cable Coupling
http://www.national.com 4
Cable-to-cable coupling occurs when two wires or cables
are run close to each other. Figure 7 shows how cable-to-
cable coupling works. Figure 7a shows two lengths of cable
(or other conductors) that are running side-by-side. Because
any two conducting bodies have capacitance between
them, called stray capacitance, a time-varying signal in one
wire can couple via that capacitance into the other wire.
That is referred to as capacitive coupling. This stray capaci-
tance, as shown in Figure 7c makes the two cables behave
as if there were a coupling capacitor between them.
Another mechanism of cable-to-cable coupling is mutual in- TL/DD/12857 – 10
ductance. Any wire carrying a time-varying current will de- FIGURE 7d. Transmission Line Model
velop a magnetic field around it. If a second conductor is Common-mode impedance coupling occurs when two cir-
placed near enough to that wire, that magnetic field will in- cuits share a common bus or wire. In Figure 8 we show a
duce a similar current in the second conductor. That type of circuit that is susceptible to that type of coupling. In that
coupling is called inductive coupling. Mutual inductance, as figure a TL092 op-amp and a 555 timer share a common
shown in Figure 7b, makes the cables behave as if a poorly return or ground. Since any conductor (including a PCB
wound transformer were connected between them. trace) is not ideal, that ground will have a non-zero imped-
In cable-to-cable coupling, either or both of those mecha- ance, Z. Because of that, the current, I, from pin 1 of the 555
nisms may be responsible for the existence of an interfer- will cause a noise voltage, VN, to develop; that voltage is
ence condition. Though there is no physical connection be- equal to I x Z. That noise voltage will appear in series with
tween the two cables, the properties we have just described the input to the op-amp. If that voltage is of sufficient ampli-
make it possible for the signal on one cable to be coupled to tude, a noise condition will result.
the other. These effects, when combined are known as the
transmission line effect and have been studied and de-
scribed thoroughly in other texts.
The basic transmission line equations are derived for distrib-
uted parameters R (series resistance), G (shunt conduct-
ance), L (series inductance), and C (shunt capacitance), all
defined per unit length of line.
Either or both of the above-mentioned properties cause the
cables to be electromagnetically coupled such that a time-
varying signal present on one will cause a portion of that
signal to appear on the other. The ‘‘efficiency’’ of the cou- TL/DD/12857 – 11
pling increases with frequency and inversely with the dis- FIGURE 8. Common-Mode Impedance Coupling
tance between the two cables. One example of cable-to-ca- While not all inclusive, these coupling paths account for,
ble coupling is telephone ‘‘crosstalk’’, in which several perhaps, 98% of all intra-system EMI situations.
phone conversations can be overheard at once. The term
crosstalk is now commonly used to describe all types of 4.0 SOURCE OF NOISE
cable-to-cable coupling. We will look at sources of EMI which involve components
that may conduct or radiate electromagnetic energy. These
sources, (component emitters), are different from the equip-
ment and subsystems we have been talking about. Compo-
nent emitters are sources of EMI which emanate from a
single element rather than a combination of components
such as was previously described. Actually, these compo-
TL/DD/12857 – 7 nent emitters require energy and connecting wires from oth-
FIGURE 7a. Parallel Conductors er sources to function. Therefore, they are not true sources
of EMI, but are EMI transducers. They convert electrical en-
ergy to electromagnetic noise.
4.1 Cables and Connectors
The three main concerns regarding the EMI role of cables
are conceptualized in Figure 9. They act as (1) radiated
TL/DD/12857 – 9 emission antennae, (2) radiated susceptibility antennae, and
FIGURE 7b. Inductive Coupling (3) cable-to-cable or crosstalk couplers. Usually, whatever is
done to harden a cable against radiated emission will also
work in reverse for controlling EMI radiated susceptibility.
The reason for this is usually, when differential-mode radiat-
ed emission or susceptibility is the failure mode, twisting
leads and shielding cables reduces EMI. If the failure mech-
anism is due to common-mode currents circulating in the
TL/DD/12857 – 8 cable, twisting leads has essentially no effect on the rela-
FIGURE 7c. Capacitive Coupling
5 http://www.national.com
tionship between each conductor and the common-mode Connectors and cables should be viewed as a system to
reference. Cable shields may help or aggravate EMI de- cost-effectively control EMI rather than to consider the role
pending upon the value of the transfer impedance of the of each separately, even though each offers specific inter-
cable shield. Transfer impedance is a figure of merit of the ference control opportunities.
quality of cable shield performance defined as the ratio of
4.2 Components
coupled voltage to surface current in X/meter. A good ca-
ble shield will have a low transfer impedance. The effective- Under conditions of forward bias, a semiconductor stores a
ness of the shield also depends on whether or not the certain amount of charge in the depletion region. If the di-
shield is terminated and, if so, how it is terminated. ode is then reverse-biased, it conducts heavily in the re-
verse direction until all of the stored charge has been re-
Connectors usually are needed to terminate cables. When
moved as shown in Figure 10. The duration, amplitude, and
no cable shields or connector filters or absorbers are used,
configuration of the recovery-time pulse (also called switch-
connectors play essentially no role in controlling EMI. The
ing time or period) is a function of the diode characteristics
influence of connector types, however, can play a major role
and circuit parameters. These current spikes generate a
in the control of EMI above a few MHz. This applies espe-
broad spectrum of conducted transient emissions. Diodes
cially when connectors must terminate a cable shield and/
with mechanical imperfections may generate noise when
or contain lossy ferrites or filter-pins.
physically agitated. Such diodes may not cause trouble if
used in a vibration-free environment.
TL/DD/12857 – 12
FIGURE 9. Cables and Connectors
TL/DD/12857 – 13
FIGURE 10. Diode Recovery Periods and Spikes
http://www.national.com 6
4.3 Power Supply Noise
Power-supply spiking is perhaps the most important contrib-
utor to system noise. When any element switches logic
states, it generates a current spike that produces a voltage
transient. If these transients become too large, they can
cause logic errors because the supply voltage drop upsets
internal logic, or because a supply spike on one circuit’s
output feeds an extraneous noise voltage into the next de-
vice’s input.
With CMOS logic in its quiescent state, essentially no cur-
rent flows between VCC and ground. But when an internal
gate or an output buffer switches state, a momentary cur-
rent flows from VCC to ground. The switching transient
caused by an unloaded output changing state typically
equals 20 mA peak. Using the circuit shown in Figure 11, TL/DD/12857 – 15
you can measure and display these switching transients un- FIGURE 12a. VIN, VCC, ICC for CL e 0 pF
der different load conditions.
Figure 12a shows the current and voltage spikes resulting
from switching a single unloaded (CL e 0 in Figure 11 )
NAND gate. These current spikes, seen at the switching
edges of the signal on VIN, increase when the output is
loaded. Figures 12b, 12c, and 12d show the switching tran-
sients when the load capacitance, CL, is 15 pF, 50 pF, and
100 pF, respectively. The large amount of ringing results
from the test circuit’s transmission line effects. This ringing
occurs partly because the CMOS gate switches from a very
high impedance to a very low one and back again. Even for
medium-size loads, load capacitance current becomes a
major current contributor.
TL/DD/12857 – 16
FIGURE 12b. VIN, VCC, ICC for CL e 15 pF
TL/DD/12857 – 17
FIGURE 12c. VIN, VCC, ICC for CL e 50 pF
TL/DD/12857 – 14
FIGURE 11. Test Circuit
TL/DD/12857 – 18
FIGURE 12d. VIN, VCC, ICC for CL e 100 pF
7 http://www.national.com
Although on standard CMOS logic internal gates generate low logic levels. ACT CMOS logic families have noise immu-
current spikes when switching, the bulk of a spike’s current nity of 2.9V for high logic levels and 700 mV for low logic
comes from output circuit transitions. Figure 13 shows the levels.
ICC current for a NAND gate, as shown in the test circuit, TABLE III. Logic Family Comparisons
with one input switching and the other at ground resulting in
no output transitions. Note the very small power-supply LS/
Charac-
glitches provoked by the input-circuit transitions. Symbol ALS HCMOS AC ACT
teristic
/TTL
Input VIH (Min) 2.0V 3.15V 3.15V 2.0V
Voltage V (Max) 0.8V 0.9V 1.35V 0.8V
IL
(Limits)
Output VOH (Min) 2.7V VCCb0.1V VCCb0.1V VCCb0.1V
Voltage V
OL (Max) 0.5V 0.1V 0.1V 0.1V
(Limits)
To illustrate noise margin and immunity, Figures 14b and
14c show the output that results when you apply several
types of simulated noise to a 74HC00’s input, Figure 14a .
Typically, even 2V or more input noise produces little
change in the output. The top trace shows noise present on
the input logic level signal and the bottom trace shows re-
sultant noise induced on the output logic level signal.
TL/DD/12857–19
FIGURE 13a. Test Circuit
TL/DD/12857 – 23
FIGURE 14a. Test Circuit
TL/DD/12857–20
FIGURE 13b. VIN, VCC, ICC TL/DD/12857 – 24
FIGURE 14b. VIN, VOUT
4.4 High Speed CMOS Logic Switching
The magnitude of noise which can be tolerated in a system
relates directly to the worst case noise immunity specified
for the logic family. Noise immunity can be described as a
device’s ability to prevent noise on its input from being
transferred to its output. It is the difference between the
worst case output levels (VOH and VOL) of the driving circuit
and the worst case input voltage requirements (VIH and VIL,
respectively) of the receiving circuit.
Using Table III as a guide, it can be seen that for TTL (LS or
ALS) devices the worst case noise immunity is typically
700 mV for the high logic level and 300 mV for the low logic
level. For HCMOS devices the worst case noise immunity is
typically 1.75V for high logic levels and 800 mV for low logic
levels. AC high speed CMOS logic families have noise im-
TL/DD/12857 – 25
munity of 1.75V for high logic levels and 1.25V for
FIGURE 14c. VIN, VOUT
http://www.national.com 8
Figure 15b shows how noise affects 74HC74’s clock input. sities increase. Our discussion on cable-to-cable coupling
Again, no logic errors occur with 2V or more of noise on the described crosstalk as appearing due to the distributed ca-
clock input. pacitive coupling and the distributed inductive coupling be-
tween two signal lines. When crosstalk is measured on an
undriven sense line next to a driven line (both terminated at
their characteristic impedances), the near end crosstalk and
the far end crosstalk have quite distinct features, as shown
in Figure 16. It should be noted that the near end compo-
nent reduces to zero at the far end and vice versa. At any
point in between, the crosstalk is a fractional sum of the
near and far end crosstalk waveforms as shown in the fig-
ure. It also can be noted that the far end crosstalk can have
either polarity whereas the near end crosstalk always has
TL/DD/12857 – 21 the same polarity as the signal causing it.
FIGURE 15a. Test Circuit The amplitude of the noise generated on the undriven
sense line is directly related to the edge rates of the signal
on the driven line. The amplitude is also directly related to
the proximity of the two lines. This is factored into the cou-
pling constants KNE and KFE by terms that include the dis-
tributed capacitance per unit length, the distributed induc-
tance per unit length, and the length of the line. The lead-to-
lead capacitance and mutual inductance thus created caus-
es ‘‘noise’’ voltages to appear when adjacent signal paths
switch.
Several useful observations that apply to a general case
can then be made:
# The crosstalk always scales with the signal amplitude VI.
# Absolute crosstalk amplitude is proportional to slew rate
TL/DD/12857 – 22 VI/tr not just 1/tr.
FIGURE 15b. VIN, VOUT
# Far end crosstalk width is always tr.
When using high speed CMOS, even with its greater noise # For tr kk 2 TL, where tr is the transition time of the
immunity, crosstalk, induced supply noise and noise tran- signal on the driven line and TL is the propagation or bus
sients become factors. Higher speeds allow the device to delay down the line, the near end crosstalk amplitude,
respond more quickly to externally induced noise transients VNE, expressed as a fraction of signal amplitude, VI, is
and accentuate the parasitic interconnection inductances KNE which is a function of physical layout only.
and capacitances that increase self-induced noise and # The high the value of ‘‘tr’’ (slower transition times) the
crosstalk. lower the percentage of crosstalk (relative to signal am-
plitude).
4.5 Signal Crosstalk
The problem of crosstalk, and how to deal with it, is becom-
ing more important as system performance and board den-
TL/DD/12857 – 26
FIGURE 16. Crosstalk
9 http://www.national.com
Although all circuit conductors have transmission line prop- ability increase in advanced logic families, the effects of
erties, these characteristics become significant when the these intrinsic electrical characteristics become more pro-
edge rates of the drivers are equal to or less than about nounced. One of these parasitic electrical characteristics is
three times the propagation delay of the line. Significant the inductance found in all leadframe materials.
transmission line properties may be exhibited, for example, Figure 17 shows a simple circuit model for a CMOS device
where devices having edge rates of 3 ns (or less) are used in a leadframe driving a standard test load. The inductor L1
to drive traces of 8 inches or greater, assuming propagation represents the parasitic inductance in the ground lead of the
delays of 1.7 ns/ft for an unloaded printed circuit trace. package; inductor L2 represents the parasitic inductance in
4.6 Signal Interconnects the power lead of the package; inductor L3 represents the
parasitic inductance in the output lead of the package; the
Of the many properties of transmission lines, two are of
resistor R1 represents the output impedance of the device
major interest to the system designer: ZOE, the effective
output, and the capacitor and resistor CL and RL represent
equivalent impedance of the line, and tpde, the effective
the standard test load on the output of the device.
propagation delay down the line. It should be noted that the
intrinsic values of line impedance and propagation delay, ZO The three waveforms shown represent how ground bounce
and tpd, are geometry-dependent. Once the intrinsic values is generated. The top waveform shows the voltage (V)
are known, the effects of gate loading can be calculated. across the load as it is switched from a logic HIGH to a logic
The loaded values for ZOE and tpde can be calculated with: LOW. The output slew rate is dependent upon the charac-
teristics of the output transistor, and the inductors L1 and
Zo
ZOE e L3, and CL, the load capacitance. In order to change the
01 a Ct/Cl output from a HIGH to a LOW, current must flow to dis-
charge the load capacitance. The second waveform shows
tpde e tpd 01 a Ct/Cl
the current that is generated as the capacitor discharges
where Ci e intrinsic line capacitance [I e bCL*(dV/dt)]. This current, as it changes, causes a
Ct e additional capacitance due to gate loading. voltage to be generated across the inductances in the cir-
These formulas indicate that the loading of lines decreases cuit. The formula for the voltage across an inductor is
the effective impedance of the line and increases the propa- V e L(dI/dt). The third waveform shows the voltage that is
gation delay. As was mentioned earlier, lines that have a induced across the inductance in the ground lead due to the
propagation delay greater than one third the rise time of the changing currents [VGB e L1*(dI/dt)]. This induced voltage
signal driver should be evaluated for transmission line ef- creates what is known as ground bounce.
fects. When performing transmission line analysis on a bus, Because the inductor is between the external system
only the longest, most heavily loaded and the shortest, least ground and the internal device ground, the induced voltage
loaded lines need to be analyzed. All lines in a bus should causes the internal ground to be at a different potential than
be terminated equally; if one line requires termination, all the external ground. This shift in potential causes the device
lines in the bus should be terminated. This will ensure simi- inputs and outputs to behave differently than expected be-
lar signal quality on all of the lines. cause they are referenced to the internal device ground,
while the devices which are either driving into the inputs or
4.7 Ground Bounce
being driven by the outputs are referenced to the their own
Ground bounce occurs as a result of the intrinsic character- device ground. External to the device, ground bounce caus-
istics of the leadframes and bond wires of the packages es input thresholds to shift and output levels to change.
used to house CMOS devices. As edge rates and drive cap-
TL/DD/12857 – 27
TL/DD/12857 – 28
FIGURE 17a. Test Circuit for Ground Bounce
FIGURE 17b. VIN, VCC, ICC
http://www.national.com 10
TL/DD/12857 – 29
FIGURE 18. Intra-System EMI Manifestations
Although this discussion is limited to ground bounce gener- might be the primary emphasis, the potential problems as-
ated during HIGH-to-LOW transitions, it should be noted sociated with either (1) susceptibility to outside conducted
that the ground bounce is also generated during LOW-to- and/or radiated emissions or (2) tendency to pollute the
HIGH transitions. This ground bounce though, has a much outside world from its own undesired emissions, come un-
smaller amplitude and therefore does not present the same der the primary classification of intra-system EMI. Corre-
concern. sponding EMI-control techniques, however, address them-
There are many factors which affect the amplitude of the selves to both self-jamming and emission/susceptibility in
ground bounce. Included are: accordance with applicable EMI specifications. The tech-
niques that will be discussed include filtering, shielding, wir-
# Number of outputs switching simultaneously: more out-
ing and grounding.
puts results in more ground bounce.
Inter-system EMI distinguishes itself by interference be-
# Type of output load: capacitive loads generate two to
tween two or more discrete and separate systems or plat-
three times more ground bounce than typical system
forms which are frequently under independent user control.
traces. Increasing the capacitive load to approximately
Culprit emissions and/or susceptibility situations are divided
60 pF –70 pF, increases ground bounce. Beyond 70 pF,
into two classes: (1) antenna entry/exit and (2) back-door
ground bounce drops off due to the filtering effect of the
entry/exit. More than 95% of inter-system EMI problems
load itself. Moving the load away from the output also
involve the antenna entry/exit route of EMI. We can group
reduces the ground bounce.
inter-system EMI-control techniques by four fundamental
# Location of the output pin: outputs closer to the ground categories: frequency management, time management, lo-
pin exhibit less ground bounce than those further away cation management and direction management.
due to effectively lower L1 and L3.
The first step in determining a solution is to identify the
# Voltage: lowering VCC reduces ground bounce. problem as either inter-system or intra-system EMI. Gener-
Ground bounce produces several symptoms: ally, if the specimen has an antenna and the problem devel-
# Altered device states. ops from what exits or enters the antenna from another
specimen or ambient, then the problem is identified as inter-
# Propagation delay degradation.
system EMI. Otherwise, it is intra-system EMI which we will
# Undershoot on active outputs. The worst-case under- discuss now.
shoot will be approximately equal to the worst-case quiet
output noise. 5.1 Intra-System EMI Control Techniques
11 http://www.national.com
TABLE IV. Effective Shielding for Different Materials
Shielding Effectiveness dB
Shielding Material Surface Resistance (X/square)
At 10 MHz At 100 MHz At 1 GHz
Arc-Sprayed Zinc 0.002 106 92 98
Silver Acrylic Paint 0.004 67 93 97
Silver Deposition 0.05 57 82 89
Silver Epoxy Paint 0.1 59 81 87
Nickel Composite 3.0 35 47 57
Carbon Composite 10.0 27 35 41
Wire Screen (0.64 mm Grid) N.A. 86 66 48
Effectiveness of shielding materials with 25 mm thickness and for frequencies for which the largest dimension of the shielding plate is less than a quarter of a
wavelength.
http://www.national.com 12
age difference between the two points. If possible, connect- pickup of differential-mode radiation. In a balanced trans-
ing both pieces of equipment to a single-point ground elimi- mission line, this is done by use of twisted-wire pairs and a
nates this voltage. Another remedy is to increase the high shielded cable.
frequency impedance along a loop that includes the path For crosstalk, the fourth coupling path, the reduction of ca-
between the ground connections of the two boxes. Exam- pacitive coupling, can be achieved by the implementation of
ples include the isolation of printed-circuit boards from their at least one of these steps:
cabinet or case, the use of a shielded isolation transformer
# Reducing the spacing between wire pairs in either or
in the signal path, or the insertion of an inductor between
both of the transmission lines.
one or both boxes and the ground conductor. The use of
balanced circuits, differential line drivers and receivers, and # Increasing the separation between the two transmission
absorbing ferrite beads and rods on the interconnecting ca- lines.
ble can further reduce currents produced by this undesir- # Reducing the frequency of operation of the source, if
able coupling path. possible.
# Adding a cable shield over either or both transmission
lines.
# Twisting the source’s or receptor’s wire pairs.
# Twisting both wire pairs in opposite directions.
The fifth coupling path conductively produces both com-
mon-mode and differential-mode noise pollution on the
power mains. Among several remedies that can suppress
the EMI there are filters and isolation transformers.
There are only about 50 common practical remedies that
can be used in most EMI situations. Of these, about 10
suffice in 80 percent of the situations. Most engineers are
aware of at least some of these remediesÐfor example,
TL/DD/12857 – 31 twisting wires to reduce radiation pickup.
FIGURE 20. Common Ground Impedance Coupling In order to attack the EMI problem, one can make use of the
information contained in Tables V and VI. First, decide what
A balanced circuit is configured so its two output signal coupling path has the worst EMI interference problem. From
leads are electrically symmetrical with respect to ground, as the 11 most common coupling paths listed at the top of the
the signal increases on one output, the signal on the other table, find the problem coupling path. Using the numbers
decreases. Differential line drivers produce a signal that is found in that table entry, locate the recommended remedy
electrically symmetrical with respect to ground from a sin- or remedies from the 12 common EMI fixes identified at the
gle-ended circuit in which only one lead is changing with bottom of the table. This procedure should be repeated until
respect to ground. Ferrite beads, threaded over electrical all significant coupling paths have been properly controlled
conductors, substantially attenuate electromagnetic inter- and the design goal has been met.
ference by turning radio-frequency energy into heat, which TABLE V. Electromagnetic Interference Coupling Paths
is dissipated in them.
In the second coupling path, a radiated electromagnetic Solution
Problem
field is converted into a common-mode voltage in the (See Table VI)
ground plane loop containing the interconnect cable and Radiated Field to Interconnecting Cable
both boxes. This voltage may be reduced if the loop area is 2, 7, 8, 9, 11
(Common-Mode)
trimmed.
Ferrite beads and ferrite filtered connectors applied to the Radiated Field to Interconnecting Cable
2, 5, 6
interconnecting cable can help reduce common mode cur- (Differential-Mode)
rents.
Interconnecting Cable to Radiated Field
1, 3, 9, 11
(Common-Mode)
Interconnecting Cable to Radiated Field
1, 3, 5, 6, 7
(Differential-Mode)
Cable-to-Cable Crosstalk 1, 2, 3, 4, 5,
6, 10, 11
Radiated Field to Box 12, 13
Box to Radiated Field 12, 13
Box-to-Box Radiation 12, 13
TL/DD/12857 – 32 Box-to-Box Conduction 1, 2, 7, 8, 9
FIGURE 21. Common-Mode, Radiated
Power Mains to Box Conduction 4, 11
Field-to-Cable Coupling
Box to Power Mains Conduction 4, 11
The third coupling path produces a differential-mode volt-
age that appears across the input terminals of the EMI re- Electromagnetic Interference Coupling
4
ceptor. One way of controlling this is to cancel or block the Paths
13 http://www.national.com
TABLE VI. Electromagnetic Interference Fixes 5.2.3 Location Management
Solution Complexity Cost Location management refers to EMI control by the selection
of location of the potential victim receptor with respect to all
1. Insert Filter in Signal Low Medium other emitters in the environment. In this regard, separation
Source distance between transmitters and receivers is one of the
2. Insert Filter in Signal Low Medium most significant forms of control since interfering source
Receptor emissions are reduced greatly with the distance between
them. The relative position of potentially interfering transmit-
3. Insert Filter in Power Low Medium ters to the victim receiver are also significant. If the emitting
Source source and victim receiver are shielded by obstacles, the
degree of interference would be substantially reduced.
4. Insert Filter in Power Low Medium
Receptor 5.2.4 Direction Management
5. Twist Wire Pair Low Low Direction management refers to the technique of EMI con-
trol by gainfully using the direction and attitude of arrival of
6. Shield Cable Low Medium electromagnetic signals with respect to the potential vic-
7. Use Balanced Circuits Medium Medium tim’s receiving antennae.
http://www.national.com 14
than 90% PCB trace dimensions plus IC package dimen- On-chip level shifting circuitry is inserted as interface be-
sions and less than 10% IC chip dimensions, which both tween the nucleus and the perimeter. It permits the nucleus
factor into the antenna gain. Therefore the IC package de- to operate at a lower VCC than the perimeter. This permits
sign should obey several rules, see Table VIII. First a ground the nucleus VCC to come through the on-chip ICC choke
pin (current return) adjacent to each power pin will minimize emerging as a bouncing VCC.
the loop sizes and simplify layout of decoupling circuits. In general, outputs toggle much less frequently than the
Also an output pin which sources or sinks a large and fre- clock. This rate is limited by software execution rates. How-
quent di/dt (for example crystal or clock pins) should also ever, for the case of a clock or address/data bus outputs
be adjacent to the power/ground pin pair to keep its current this may not be the case. Looking at time domain EMI in-
loops small. It helps to have small package dimensions. A stead of frequency domain EMI (which may have some time
last point is that sharing of a common power/ground should averaging) may motivate one to take action and reduce EMI
be limited to prevent their voltage from bouncing due to the from the chip’s output drivers. CMOS output drivers turn on
total Ldi/dt a iR. This bouncing is conducted to inputs shar- fast, in about 1 ns. This causes 2 problems. The first is
ing the common power/ground which in several cases are called ‘‘shoot through’’ current which flows from DVCC to
susceptible since their VIH and VIL trip points are referenced DGND when both the pull up and the pull down drivers are
to it. The bouncing is conducted through non-tristated out- on and the output load is small. The second is due to output
puts sharing this common power/ground and is passed on current magnitude increasing to maximum in about 1 ns
outside the IC as additional conducted and radiated emis- when the output changes state. Both problems are solved
sions from bouncing VOH or VOL levels. by using fast turn off, gradual turn on device drivers.
TABLE VIII. Package and PCB Traces form Loop or Figure 22 illustrates, in block diagram form, how the chip is
Whip Antennae which Radiate partitioned for EMC. Note that the level shifter block is multi-
channel, and most channels need dedicated level shifting
1. Ground pin (current return) not adjacent to power devices in one direction only as shown. The power/ground
pin). Pair should be together. pads named DVCC (Driver VCC), LVCC (Logic VCC), VCC (a
2. Power/Ground pair not adjacent to output pin global VCC to most of the chip), GND (a global GND to most
(especially crystal of clock pins). of the chip) and DGND (driver ground) are all available. To
minimize package pin count, DVCC and LVCC can be bond-
3. Too much sharing of common power/ground ed to the same pin, GND and DGND can be bonded to the
having voltage bounce due to its total Ldi/dt a iR. same pin, and the VCC pad is not bonded. A higher pin
a. Conducted emissions to susceptible I/O count package is used to bring out each pad to a separate
b. Radiated emissions from susceptible I/O. pin to allow study of the several packaging options. The VCC
pad is not bonded when utilizing the on-chip ICC choke to
4. Large package dimensions. prevent RE (radiated emissions) from VCC bounce.
Three operating modes can be evaluated with GND and
6.2 Silicon Design Changes DGND pads always connected together to 0V. The first
National’s design goal was to reshape the IC pin current mode applies 5V to DVCC, LVCC and VCC to check the RE
waveforms from their original values as follows: ICC pulses improvements due to gradual turn on outputs alone. The
to 3X rise time, 0.2X amplitude and 5X width; IOL and IOH second mode applies 5V to DVCC and LVCC where VCC
pulses to 10X rise time. A second goal is to provide sepa- floats as a test point. This checks the RE improvements due
rate power and ground bus systems on-chip for each of the to gradual turn on and ICC choking. The third mode applies
following groups: Chip digital logic, Chip I/O buffers, and if 5V to DVCC and 3V to VCC and LVCC to check the RE
present a chip A/D converter (or other analog intensive sec- improvement in this mode which actually requires a VCC
tions). This is an extension of proper IC packaging pinout additional package pin.
covered earlier to prevent conducted emissions to suscepti- TABLE IX. Results of Various Pinout Configurations
ble circuitry.
Config- Config- Config-
Inspection of these goals leads us to divide the chip into Chip Pads
uration A uration B uration C
separate areas which we named the nucleus, the perimeter
and the A/D. Also the partitioning of the nucleus (containing DVCC 5V 5V 5V
the digital logic functions) and the perimeter (containing dig-
ital I/O, oscillator and comparator functions) allows use of LVCC 5V 5V 3V
an on-chip device to choke the nucleus’ ICC current, thereby VCC 5V Test Point 3V
wave shaping the ICC. The ICC choke, however, causes the
nucleus VCC voltage to dip by 0.5V to 1V at each clock GND 0 0 0
edge. These VCC voltage swings will be confined to within DGND 0 0 0
the chip’s nucleus where the circuitry (digital logic) is toler-
ant and no significant radiation antennae exist. EMI Baseline b 20 dB b 6 dB
15 http://www.national.com
TL/DD/12857 – 33
FIGURE 22. Block Diagram of EMC Circuitry
http://www.national.com 16
Test results in Figure 26 show emissions of 45 dBmV re-
duced to less than 20 dBmV on the same device. Unchoked
tests were performed using the bonding options previously
discussed.
Figure 25 gives the predicted performance degradation from
implementing gradual turn-on outputs. High to low transition
time propagation delays are doubled and low to high tran-
sitions are slowed by about 50%. Remember, though, that
propagation delays for an embedded microcontroller are
usually of minimal concern since control is based on levels
or edges of individual signals or on timing which is under
software control. Software control is slow enough that the
change in hardware timing is probably negligible.
TL/DD/12857 – 35
FIGURE 24a. Original ICC An indication of the type of emission reduction resulting
from the implementation of gradual turn-on can be seen if
you look at the baselines of the plots of Figure 26. Changes
of the outputs occur at low frequency with large amplitude,
fast rise time current spikes. The low frequency of the signal
keeps the harmonic close together and makes the emis-
sions appear broadband. The fast rise time maintains a
higher level of emissions to a higher frequency.
TL/DD/12857 – 36
FIGURE 24b. ICC with Improvements
TL/DD/12857 – 39
FIGURE 26a. VOL
TL/DD/12857 – 37
FIGURE 25a. Emissions without VCC Choke
TL/DD/12857 – 40
FIGURE 26b. VOH
TL/DD/12857 – 38
FIGURE 25b. Emissions with Addition of VCC Choke
17 http://www.national.com
7.0 SYSTEM PERFORMANCE USING EMI REDUCED order to achieve a predetermined signal to noise ratio
PARTS (SNR) at the speaker. In this case the input was chosen
We have discussed in previous sections that we can reduce to be a 22 kHz deviation of a frequency modulated signal
EMI by proper board design and by the selection of low EMI within the US commercial FM radio broadcast band. The
devices. We will now investigate whether EMI reduced parts signal to noise criterion used was 30 dB.
are worth it. The goal of this testing is to determine the 2. Two other devices, a COP888GG and a COP87M88GG
relative electromagnetic emissions from various production OTP, were placed in a common two sided PCB in a TEM
options of the COP8 microcontroller. Since most people use cell and were programmed to run a standard EMI test
OTPs to qualify their end product for emissions approval, program included in National’s first silicon engineering
both OTPs and ROMmed devices will be tested. The op- code. Load capacitors of 10 pF and 4.7 kX pull-up resis-
tions chosen were: tors used on the eight pins of Port L. The port was pro-
1. COP8788EGMH (OTP) part used for development and grammed to count in a binary manner with the least sig-
Prototype approval nificant bit changing state every 100 ms. (Figure 27 illus-
trates the board layout.) Four AA batteries were attached
2. COP888EK (masked ROM) part used for final production
to the board and the supply voltage was regulated with a
(EMI reduced)
Zener diode.
3. COP8788GG (OTP) part used for development and Pro-
Emissions were measured with the TEM cell on a Tektro-
totype approval
nix Model 2754P Spectrum Analyzer over the range of
4. COP888GG (masked ROM) part used for final production DC to 750 MHz.
(EMI reduced)
7.2 Radio Results
7.1 Test Procedure
Worst case useable sensitivity is improved by about 17 dB
These tests were performed in two ways: (25 dBmV to 8 dBmV) by the change from the
1. Two DIP parts, a COP888EK and a COP888EGMH OTP COP888EGMH OTP to the mask programmed COP888EK.
device, were programmed for use as the main processor Of course not all frequencies show this improvement, but
in an automotive radio and the relative performance of there is some increased sensitivity throughout the spectrum.
the radio was measured for each device. The measure of Note that two local radio stations (at 105.7 MHz and 106.5
performance chosen was Usable Sensitivity. This is de- MHz) were received even without an external antenna.
fined as the minimum input required at the antenna in
TL/DD/12857 – 42
FIGURE 27. PCB Layout for EMI Testing
http://www.national.com 18
TL/DD/12857 – 41
FIGURE 28. Usable Sensitivity Comparison
7.3 EMI Board Results Much more noticeable is the reduction of spectral density in
Except for the fundamental oscillator frequency of approxi- the noise. Off harmonic noise is not noticeable.
mately 5 MHz, emissions are down at least 10 dB on the
monolithic masked device over the multichip OTP device.
Emissions beyond 100 MHz are practically eliminated.
19 http://www.national.com
TL/DD/12857 – 43
FIGURE 29. COP888GG with 5 MHz Crystal Oscillator
TL/DD/12857 – 44
FIGURE 30. COP87L88GG with 5 MHz Crystal Oscillator
http://www.national.com 20
7.4 Conclusions The type of logic to be used is normally an early design
Radiated Emissions do not come from a device, but from a decision, so that control of edge speeds and, hence, emis-
board. Devices have some control over the emissions be- sions and susceptibiity is made early. Of course, other fac-
cause they generate the noise which is radiated. The equa- tors such as required system performance, speed, and tim-
tion normally used to predict emission from a system is: ing considerations must enter into this decision. If possible,
2 (/2 design the circuit with as slow speed logic as possible. The
1.32 c 10b3 cI cA cFreq 2
Ð # 2qD J (
l
lE lMax e D
c 1a use of slow speed logic, however, does not guarantee that
EMC will exist when the circuit is built; so proper EMC tech-
where, niques should still be implemented consistently during the
lElMax is the maximum E-field in the plane of the loop in remainder of the circuit design.
mV/m 8.2 Component Layout
l is the current amplitude, at the frequency of interest, in Component layout is the second stage in PCB design. Sche-
milliamps matics tell little or nothing about how systems will perform
A is the loop area is cm2 once the board is etched, stuffed, and powered. A circuit
l is the wavelength at the frequency of interest schematic is useful to the design engineer, but an experi-
D is the observation distance in meters enced EMC engineer refers to the PCB when trouble-shoot-
ing. By controlling the board layout in the design stage, the
Freq is the frequency in MHz
designer realizes two benefits: (1) a decrease in EMI prob-
and the perimeter of the loop P kk l lems when the circuit or system is sent for EMI or quality
When applied to a chip, the area is very small and current assurance testing; and (2) the number of EMI coupling
is relatively low compared to current consumption in a paths is reduced, saving troubleshooting time and effort lat-
typical system, thus the emissions are low. er on.
8.0 DESIGN GUIDELINES Some layout guidelines for arranging components according
to logic speed, frequency, and function are shown in Figure
The growth of concern over electromagnetic compatibility
31. These guidelines are very general. A particular circuit is
(EMC) in electronic systems continues to rise in the years
likely to require a combination and/or trade-offs of these
since the FCC proclaimed that there shall be no more
arrangements. Isolation of the I/O from digital circuitry is
pollution of the electromagnetic spectrum. Still, designers
important where emissions or susceptibility may be a prob-
have not yet fully come to grips with a major source and
lem. For the case of emissions, a frequently encountered
victim of electromagnetic interferenceÐthe PCB. The
coupling path involves digital energy coupling through I/O
most critical stage for addressing EMI is during the circuit
circuitry and signal traces onto I/O cables and wires, where
board design. Numerous tales of woe can be recounted
the latter subsequently radiate. When susceptibility is a
about the eleventh hour attempt at solving an EMI prob-
problem, it is common for the EMI energy to couple from
lem by retrofit because EMC was given no attention dur-
I/O circuits onto sensitive digital lines, even though the I/O
ing design. This retrofit ultimately costs much more than
lines may be ‘‘opto-coupled’’ or otherwise supposedly iso-
design stage EMC, holds up production and generally
lated. In both situations, the solution often lies in the proper
makes managers unhappy. With these facts in mind, let’s
electrical and physical isolation of analog and low speed
address electromagnetic compatibility considerations in
digital lines from high speed circuits. When high speed sig-
PCB design.
nals are designed to leave the board, the reduction of EMI is
8.1 Logic Selection usually performed via shielding of I/O cables and is not con-
Logic selection can ultimately dictate how much attention sidered here.
must be given to EMC in the total circuit design. The first Therefore, a major guideline in layout out boards is to iso-
guideline should be: use the slowest speed logic that will late the I/O circuitry from the high speed logic. This method
do the job. Logic speed refers to transition times of output applies even if the logic is being clocked at ‘‘only’’ a few
signals and gate responses to input signals. Many emis- MHz. Often, the fundamental frequency is of marginal inter-
sions and susceptibility problems can be minimized if a est, with the harmonics generated from switching edges of
slow speed logic is used. For example, a square wave the clock being the biggest emission culprits. Internal sys-
clock or signal pulse with a 3 ns rise time generates radio tem input/output PCB circuitry should be mounted as close
frequency (100 MHz and higher) energy that is gated to the edge connector as possible and capacitive filtering of
about on the PCB. It also means that the logic can re- these lines may be necessary to reduce EMI on the lines.
spond to comparable radio frequency energy if it gets High speed logic components should be grouped together.
onto the boards. Digital interface circuitry and I/O circuitry should be physi-
cally isolated from each other and routed on separate con-
nectors, if possible as shown in Figure 27d.
21 http://www.national.com
8.3 Power Supply Bussing
Power supply bussing is the next major concern in the de-
sign phase. Isolated digital and analog power supplies must
be used when mixing analog and digital circuitry on a board.
The design preferably should provide for separate power
supply distribution for both the analog and digital circuitry.
Single point common grounding of analog and digital power
supplies should be performed at one point and one point
onlyÐusually at the motherboard power supply input for
TL/DD/12857–45 multi-card designs, or at the power supply input edge con-
FIGURE 31a. Board Layout nector on a single card system. The fundamental feature of
good power supply bussing, however, is low impedance and
good decoupling over a large range of frequencies. A low
impedance distribution system requires two design features:
(1) proper power supply and return trace layout and (2)
proper use of decoupling capacitors.
An exception to the rule of analog and digital ground sepa-
ration applies in the case of high speed interface between
analog and digital portions. If possible, the ground plane
should be continuous under these interface traces to im-
prove the proximity of the return trace and thus minimize the
TL/DD/12857–46 size of the loop.
FIGURE 31b. Board Layout At high frequencies, PCB traces and the power supply bus-
ses ( a VCC and 0V) are viewed as transmission lines with
associated characteristic impedance, ZO, as modeled in
Figure 28. The goal of the designer is to maximize the ca-
pacitance between the lines and minimize the self-induc-
tance, thus creating a low ZO. Table X, shows the character-
istic impedance of various two-trace configurations as a
function of trace width, W, and trace separation, h.
TL/DD/12857–47
FIGURE 31c. Board Layout
TL/DD/12857 – 49
FIGURE 32. Transmission Line Modelling of PCB Traces
TL/DD/12857–48
FIGURE 31d. Board Layout
http://www.national.com 22
TABLE X. Characteristic Impedance of Any one of the three configurations may be viewed as a
Different Conductor Pairs possible method of routing power supply (or signal) traces.
The most important feature of this table is the noticeable
difference in impedance between the parallel strips and
strip over ground plane compared with the side-by-side con-
figurations.
As an example of the amount of voltage that can be gener-
ated across the impedance of a power bus, consider TTL
logic which pulls a current of approximately 16 mA from a
Strips Over supply that has a 25X bus impedance (this assumes no
W/h or Parallel Strips Side decoupling present). The transient voltage is approximately
Ground
D/W Strips by Side dV e 0.016 x 25X e 400 mV, which is equal to the noise
Plane immunity level of the TTL logic. A 25X (or higher) imped-
0.5 377 377 N/A ance is not uncommon in many designs where the supply
and return traces are routed on the same side of the board
0.6 281 281 N/A in a side-by-side fashion. In fact, it is not uncommon to find
0.7 241 241 N/A situations where the power supply and return traces are
routed quite a distance from each other, thereby increasing
0.8 211 211 N/A the overall impedance of the distribution system and the
0.9 187 187 N/A size of the loop antenna. This is obviously a poor layout.
Power and ground planes offer the lowest overall imped-
1.0 169 169 0
ance. The use of these planes leads the designer closer to
1.1 153 153 25 a multi-layer board. At the very least, it is recommended that
all open areas on the PCB be ‘‘landfilled’’ with an intercon-
1.2 140 140 34 nected 0V reference plane so that ground impedance is
1.5 112 112 53 minimized. This landfill technique may be ineffective, how-
ever if minimum width traces connect the filled areas.
1.7 99 99 62
Multi-layer boards offer a considerable reduction in power
2.0 84 84 73 supply impedance, as well as other benefits. Extending the
table we just saw, the impedance of a multi-layer power/
2.5 67 67 87
ground plane bus grows very small (on the order of an X or
3.0 56 56 98 less), assuming a W/h ratio greater than 100. Multi-layer
board designs also pay dividends in terms of greatly re-
3.5 48 48 107
duced EMI, and they provide close control of line impedanc-
4.0 42 42 114 es where impedance matching is important. In addition,
5.0 34 34 127
6.0 28 28 137
7.0 24 24 146
8.0 21 21 153
9.0 19 19 160
10.0 17 17 166
12.0 14 14 176
15.0 11.2 11.2 188
20.0 8.4 8.4 204
25.0 6.7 6.7 217
30.0 5.6 5.6 227
40.0 4.2 4.2 243
50.0 3.4 3.4 255
23 http://www.national.com
TL/DD/12857 – 53
FIGURE 33. Decoupling Example
http://www.national.com 24
It is good practice to distribute decoupling capacitors evenly
throughout the logic on the board, placing at least one ca-
pacitor for every package as close to the power and ground
pins as possible. The parasitic inductance in the capacitor
leads can be greatly reduced or eliminated by the use of
surface mount chip capacitors soldered directly onto the
board at the appropriate locations. Decoupling capacitors TL/DD/12857 – 56
need to be of the high K ceramic type with low equivalent FIGURE 35b. Parallel Termination
series resistance (ESR), consisting primarily of series induc-
tance and series resistance. Capacitors using Z5U dielectric Parallel terminations are not generally recommended for
have suitable properties and make a good choice for decou- CMOS circuits due to their power consumption, which can
pling capacitors; they offer minimum cost and effective per- exceed the power consumption of the logic itself. The pow-
formance. er consumption of parallel terminations is a function of the
8.5 Proper Signal Trace Layout resistor value and the duty cycle of the signal. In addition,
parallel termination tends to bias the output levels of the
Although crosstalk cannot be totally eliminated, there are driver towards either VCC or ground depending on which
some design techniques that can reduce system problems bus the resistor is connected to. While this feature is not
resulting from crosstalk. In any design, the distance that desirable for driving CMOS inputs because the trip levels
lines run adjacent to each other should be kept as short as are typically VCC/2, it can be useful for driving TTL inputs
possible. The best situation is when the lines are perpendic- where level shifting is desirable in order to interface with
ular to each other. Crosstalk problems can also be reduced CMOS devices.
by moving lines further apart or by inserting ground lines or
planes between them. AC parallel terminations work well for applications where
the increase in bus delays caused by series terminations are
For those situations where lines must run parallel, as in ad- undesirable. The effects of AC parallel terminations are sim-
dress and data buses, the effects of crosstalk can be mini- ilar to the effects of standard parallel terminations. The ma-
mized by line termination. Terminating a line in its character- jor difference is that the capacitor blocks any DC current
istic impedance reduces the amplitude of an initial crosstalk path and helps to reduce power consumption.
pulse by 50%. Terminating the line will also reduce the
amount of ringing. Thevenin terminations are not generally recommended due
to their power consumption. Like parallel terminations, a DC
There are several termination schemes which may be used. path to ground is created by the terminating resistors. The
They are series, parallel, AC parallel and Thevenin termina- power consumption of a Thevenin termination, though, will
tions. AC parallel and series terminations are the most use- generally be independent of the signal duty cycle. Thevenin
ful for low power applications since they do not consume terminations are more applicable for driving CMOS inputs
any DC power. Parallel and Thevenin terminations increase because they do not bias the output levels as paralleled
DC power consumption. terminations do. It should be noted that output lines with
Series terminations are most useful in high-speed applica- Thevenin terminations should not be left floating since this
tions where most of the loads are at the far end of the line. will cause the undriven input levels to float between VCC
Loads that are between the driver and the end of the line and ground, increasing power consumption.
will receive a two-step waveform. The first wave will be the
incident wave. The amplitude is dependent upon the output
impedance of the driver, the value of the series resistor and
the impedance of the line according to the formula:
VCC c ZOE
VW e
ZOE a RS a ZS
where RS is the series Resistor
ZS is the output impedance of the driver
ZOE is the equivalent line impedance TL/DD/12857 – 57
FIGURE 36a. AC Parallel Termination
TL/DD/12857 – 55
FIGURE 35a. Series Termination
25 http://www.national.com
8.6 Ground Bounce 8.8 Cables and Connectors
Observing either one of the following rules is sufficient to Several options are available to reduce EMI from a typical
avoid running into any of the problems associated with ribbon cable used to interconnect pieces of equipment.
ground bounce: These include:
# First, use caution when driving asynchronous TTL-level # Reduce spacing between conductors (h in Figure 37 ) by
inputs from CMOS outputs. Ground bounce glitches may reducing the size of wires used and reducing the insula-
cause spurious inputs that will alter the state of non- tion thickness. (This could have a negative impact on
clocked logic. crosstalk within the cable.)
# Second, caution when running control lines (set, reset, # Join alternate signal returns together at the connectors
load, clock, chip select) which are glitch-sensitive at each end of the cable.
through the same devices that drive data or address # Twist parallel wire pairs in ribbon cables.
lines.
# Shield ribbon cable with metal foil cover (superior to
When it is not possible to avoid the above conditions, there braid).
are simple precautions available which can minimize ground
# Replace discrete ribbon cable with stripline flexprint ca-
bounce noise. These are:
ble.
# Choose package outputs that are as close to the ground In the case of joining alternate signal returns, wire N is car-
pin as possible to drive asynchronous TTL-level inputs.
rying the signal current, IN, whereas its mates, N b 1 and
# Use the lowest VCC possible or separate the power sup- N a 1 wires are each carrying one half of the return cur-
plies. rents, IN b 1 and IN a 1, respectively. Thus, radiation from
# Use board design practices which reduce any additive pair N and N b 1 is out of phase with radiation from pair N
noise sources, such as crosstalk, reflections, etc. and N a 1 and will tend to cancel. In practice, however, the
net radiation is reduced by 20 dB – 30 dB with 30 dB being a
8.7 Components
good default value.
The interference effect by rectifier diodes, typically found in
The opposite of this is to conserve signal returns by only
power supply sections of PCBs, can be minimized by one or
using one, or two, wires to service N data lines in a ribbon
more of the following measures:
cable. For data lines farther from the return line, the differ-
# Placing a bypass capacitor in parallel with each rectifier ential mode radiation becomes so great that this cable
diode tends to maximize EMI radiation. Another disadvantage of
# Placing a resistor in series with each rectifier diode this approach is poor impedance control in the resulting
# Placing an R-F bypass capacitor to ground from one or transmission line. This could result in distortion of pulses
both sides of each rectifier diode and causes reflections, especially for high-speed logic, and
common return impedance noise in this single ground wire.
# Operating the rectifier diodes well below their rated cur-
rent capability Ideally, connectors should have negligible resistance for ob-
vious reasons other than EMI control. They should provide
for foolproof alignment to minimize the possibility of contact
damage over time and use which would increase the resist-
ance and be prone to vibration and shock. The required
force should be adequate to provide good mating between
contacts which will insure low resistance, but should be min-
imized to limit likelihood of damage. Connectors should
mate with little friction to minimize the effects of continual
disconnections and connections increasing the contact re-
sistance with use as the contacts wear out. A contamination
free design should be used to avoid corrosion and oxidation
increasing resistance and susceptibility to shock and vibra-
tion causing intermittent contact.
8.9 Special Considerations with Development Tools
TL/DD/12857–59 The following set of guidelines have been compiled from the
FIGURE 37a. RF Filtering of Diode experiences of the applications team at Embedded Tech-
nologies Division at National Semiconductor. They should
be considered additional techniques and guidelines to be
followed concurrently with the standard ones already pre-
sented. Some are general and some may be specific to de-
velopment systems use.
Ground bounce prevention and minimization techniques
presented in this paper should be strictly adhered to when
using ‘373 type transparent latches on any controller’s ex-
ternal address/data bus. Multiple simultaneously switching
outputs could produce ground bounce significant enough to
TL/DD/12857–60
FIGURE 37b. Connectors and Cables
http://www.national.com 26
TL/DD/12857 – 61
FIGURE 38a. Alternating Signal Return Minimizes Radiation
TL/DD/12857 – 62
FIGURE 38b. Single Signal Return Maximizes Radiation
cause false latching. Observe good EMI planning by locat- to move the problem from one point in the affected signal
ing the latches as close to the controller as possible. The waveform to a different point. Thus, apparent ‘‘noise glitch-
use of multi-layer PCBs with good ground planes and follow- es’’ that caused a latch to erroneously trigger when the in-
ing appropriate layout techniques is also essential, especial- put data was still changing, may now come at a time when
ly if emulation will be done at frequencies above 10 MHz. they are non-destructive such as at a point when the input
With the foregoing discussions about ‘‘antenna farms’’, radi- data is now stable.
ated noise and ideal connector characteristics, it becomes Some applications require driving the controller clock input,
obvious that wire-wrap boards and the use of IC sockets is CKI, with an external signal. Most emulator tools are all
absolutely out of the question. The concern here is not so clocked using a crystal network with the controller so that
much EMI affecting the outside world but EMI strangling the the generation of the system timing is contained on the tool
operation of the module itself. itself. Consequently, there is no connection between the
The inputs to the buffers in a ‘244 type octal buffer package emulator cable connector on the tool and the CKI pin at the
are placed adjacent or side-by-side outputs of other buffers controller. However, when the emulator cable is now insert-
in the package. This configuration would tend to maximize ed into the target board, the target board’s clock signal trav-
the crosstalk or noise coupling from the inputs to the out- eling along the cable couples noise onto adjacent signal
puts. On the other hand, the buffer inputs in a ‘544 type lines causing symptoms pointing to an apparent failure of
package are on one side of the package and the outputs are the emulator tool. The recommendation is to disable the
on the other. The use of these package types in high speed clock drive to the CKI pin at the controller pad on the target
designs can facilitate board layout to help reduce the ef- board whenever the emulator tool is connected. The emula-
fects of crosstalk. tor tools supply the system clock so there is no need for the
Use extra heavy ground wires between emulator and target clock on the target and signal crosstalk on the emulator
board. Rely on the ground returns in the emulator cable for cable can be greatly reduced with minimal implementation.
reduction of differential-mode noise radiated from the cable If one insists that the emulator tool and the target be syn-
but heavy-duty help is required for reducing power line im- chronous, then bring the clock signal from the target to the
pedance in the integrated development system. emulator tool external to the emulator cable via twisted wire
pair or coax cable. Remove the clock drive connection to
Unused controller inputs, most importantly any external in-
CKI at the target to prevent the signal from entering the
terrupts and RDY, must be tied to VCC (or GND) directly or
cable. Finally, remove crystal components on the emulator
through a pull-up (or pull-down) resistor. This not only tends
tool to prevent problems with the signal.
to reduce power consumption, but will avoid noise problems
triggering an unwanted action. Connecting boards and modules together to make a totally
unique system in which EMC was practiced is necessary to
In order to reduce the effects of noise generated by high
ensure little problem with the environment. But, connecting
speed signal changes, a sort of Frequency Management
an emulator tool makes it an entirely new and unique sys-
technique might be applied. If possible, develop application
tem, both in physical and electrical properties. Treat the em-
hardware and software at a slower crystal operating fre-
ulator tool as part of the system during the design phase
quency. If ringing, crosstalk, or other combination of radiat-
and development phase.
ed and conducted noise problems exists, the result may be
27 http://www.national.com
Understanding and Eliminating EMI in Microcontroller Applications
9.0 SUMMARY 4. Engineering Staff, Don White Consultants, Inc., The
The design and construction of an electromagnetically com- role of Cables & Connectors in Control of EMI , EMC
patible PCB does not necessarily require a big change in Technology, July 1982, pp 16-26
current practices. On the contrary, the implementation of 5. Fairchild Semiconductor, Design Considerations , Fair-
EMC principles during the design process can fit in with the child Advanced CMOS Technology Logic Data
ongoing design. When EMC is designed into the board, the Book, 1987, pp 4-3 to 4-12
requirements to shield circuitry, cables, and enclosures, as 6. Fairchild Semiconductor, Understanding and Minimizing
well as other costly eleventh hour surprises, will be drasti- Ground Bounce , Application Note DU-6, August 1987
cally reduced or even eliminated. Without EMC in the design
7. Interference Control Technology, Introduction to EMI/
stage, production can be held up and the cost of the project
RFI/EMC , Course Notebook on Electromagnetic
increases.
Compatibility, August 1988
REFERENCES 8. Violette, Michael F. and J.L., EMI Control in the Design
1. Alkinson, Kenn-Osburn, John-White, Donald, Taming and Layout of Printed Circuit Boards , EMC Technolo-
EMI in Microprocessor Systems , IEEE Spectrum, De- gy, March-April 1986, pp 19-32
cember 1985, pp 30-37 9. Violette, Michael F., Curing Electromagnetic Interfer-
2. Balakrishnan, R.V., Reducing Noise on Microcomputer ence , Radio-Electronics, November 1985, pp 50-56
Buses , National Semiconductor Application Note 10. Wakeman, Larry, High-speed-CMOS Designs Address
337, May 1983 Noise and I/O Levels , National Semiconductor Appli-
3. Brewer, Ron W., Test Methodology to Determine Levels cation Note 375, September 1984
of Conducted and Radiated Emissions from Computer 11. White, Donald R.J., EMI Control Methods and Tech-
Systems , EMC Technology, July 1982, p 10 niques . Electromagnetic Interference and Compati-
bilityÐVol 3, copyright 1988 by Interference Control
Technologies
NATIONAL’S PRODUCTS ARE NOT AUTHORIZED FOR USE AS CRITICAL COMPONENTS IN LIFE SUPPORT
DEVICES OR SYSTEMS WITHOUT THE EXPRESS WRITTEN APPROVAL OF THE PRESIDENT OF NATIONAL
SEMICONDUCTOR CORPORATION. As used herein:
1. Life support devices or systems are devices or 2. A critical component is any component of a life
systems which, (a) are intended for surgical implant support device or system whose failure to perform can
into the body, or (b) support or sustain life, and whose be reasonably expected to cause the failure of the life
failure to perform, when properly used in accordance support device or system, or to affect its safety or
with instructions for use provided in the labeling, can effectiveness.
be reasonably expected to result in a significant injury
to the user.
AN-1050
National does not assume any responsibility for use of any circuitry described, no circuit patent licenses are implied and National reserves the right at any time without notice to change said circuitry and specifications.
IMPORTANT NOTICE
Texas Instruments Incorporated and its subsidiaries (TI) reserve the right to make corrections, modifications, enhancements, improvements,
and other changes to its products and services at any time and to discontinue any product or service without notice. Customers should
obtain the latest relevant information before placing orders and should verify that such information is current and complete. All products are
sold subject to TI’s terms and conditions of sale supplied at the time of order acknowledgment.
TI warrants performance of its hardware products to the specifications applicable at the time of sale in accordance with TI’s standard
warranty. Testing and other quality control techniques are used to the extent TI deems necessary to support this warranty. Except where
mandated by government requirements, testing of all parameters of each product is not necessarily performed.
TI assumes no liability for applications assistance or customer product design. Customers are responsible for their products and
applications using TI components. To minimize the risks associated with customer products and applications, customers should provide
adequate design and operating safeguards.
TI does not warrant or represent that any license, either express or implied, is granted under any TI patent right, copyright, mask work right,
or other TI intellectual property right relating to any combination, machine, or process in which TI products or services are used. Information
published by TI regarding third-party products or services does not constitute a license from TI to use such products or services or a
warranty or endorsement thereof. Use of such information may require a license from a third party under the patents or other intellectual
property of the third party, or a license from TI under the patents or other intellectual property of TI.
Reproduction of TI information in TI data books or data sheets is permissible only if reproduction is without alteration and is accompanied
by all associated warranties, conditions, limitations, and notices. Reproduction of this information with alteration is an unfair and deceptive
business practice. TI is not responsible or liable for such altered documentation. Information of third parties may be subject to additional
restrictions.
Resale of TI products or services with statements different from or beyond the parameters stated by TI for that product or service voids all
express and any implied warranties for the associated TI product or service and is an unfair and deceptive business practice. TI is not
responsible or liable for any such statements.
TI products are not authorized for use in safety-critical applications (such as life support) where a failure of the TI product would reasonably
be expected to cause severe personal injury or death, unless officers of the parties have executed an agreement specifically governing
such use. Buyers represent that they have all necessary expertise in the safety and regulatory ramifications of their applications, and
acknowledge and agree that they are solely responsible for all legal, regulatory and safety-related requirements concerning their products
and any use of TI products in such safety-critical applications, notwithstanding any applications-related information or support that may be
provided by TI. Further, Buyers must fully indemnify TI and its representatives against any damages arising out of the use of TI products in
such safety-critical applications.
TI products are neither designed nor intended for use in military/aerospace applications or environments unless the TI products are
specifically designated by TI as military-grade or "enhanced plastic." Only products designated by TI as military-grade meet military
specifications. Buyers acknowledge and agree that any such use of TI products which TI has not designated as military-grade is solely at
the Buyer's risk, and that they are solely responsible for compliance with all legal and regulatory requirements in connection with such use.
TI products are neither designed nor intended for use in automotive applications or environments unless the specific TI products are
designated by TI as compliant with ISO/TS 16949 requirements. Buyers acknowledge and agree that, if they use any non-designated
products in automotive applications, TI will not be responsible for any failure to meet such requirements.
Following are URLs where you can obtain information on other Texas Instruments products and application solutions:
Products Applications
Audio www.ti.com/audio Communications and Telecom www.ti.com/communications
Amplifiers amplifier.ti.com Computers and Peripherals www.ti.com/computers
Data Converters dataconverter.ti.com Consumer Electronics www.ti.com/consumer-apps
DLP® Products www.dlp.com Energy and Lighting www.ti.com/energy
DSP dsp.ti.com Industrial www.ti.com/industrial
Clocks and Timers www.ti.com/clocks Medical www.ti.com/medical
Interface interface.ti.com Security www.ti.com/security
Logic logic.ti.com Space, Avionics and Defense www.ti.com/space-avionics-defense
Power Mgmt power.ti.com Transportation and Automotive www.ti.com/automotive
Microcontrollers microcontroller.ti.com Video and Imaging www.ti.com/video
RFID www.ti-rfid.com
OMAP Mobile Processors www.ti.com/omap
Wireless Connectivity www.ti.com/wirelessconnectivity
TI E2E Community Home Page e2e.ti.com
Mailing Address: Texas Instruments, Post Office Box 655303, Dallas, Texas 75265
Copyright © 2011, Texas Instruments Incorporated